Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

Pushkar Arande points out that 80 percent of a modern semiconductor fab's cost structure is entirely fixed. Because of this massive overhead, small percentage improvements in machine utilization and wafer yields yield nonlinear revenue boosts, explaining why rising capital expenses have consolidated leading-edge manufacturing down to three foundries.

Related Insights

While TSMC's Arizona expansion has been complex, it's already achieving yields comparable to its Taiwan facilities. An expert believes this success comes at a price, with higher costs likely being a permanent feature of US-based manufacturing.

A single year of Nvidia's revenue is greater than the last 25 years of R&D and capex from the top five semiconductor equipment companies combined. This suggests a massive 'capex overhang,' meaning the primary bottleneck for AI compute isn't the ability to build fabs, but the financial arrangements to de-risk their construction.

As the dominant chip foundry, TSMC acts as a "kingmaker" by methodically managing its capacity expansion to ensure supply always lags explosive demand. According to Semi Analysis, this strategy is intentional, as there's no incentive to "let the market go out over its skis," which maintains high prices and benefits overflow competitors like Intel.

For the next few years, the primary constraint on memory production is not a shortage of manufacturing equipment. Rather, it's the physical lack of clean room space. Memory companies, burned by years of low margins, failed to build new fabs, which have a two-year construction lead time.

New chip fab ventures face immense hurdles because fabrication is less like following a manual and more like mastering a recipe through decades of trial and error. This accumulated, non-transferable knowledge, likened to "cooking," creates a significant moat for incumbents like TSMC.

Pushkar Arande states that expanding wafer manufacturing cannot match the exponential compute demand of AI because suppliers like ASML output fewer than 100 EUV tools annually. Even with limitless capital, long manufacturing lead times and fixed machine supply chains prevent leading-edge fabs from scaling beyond roughly double current output.

The immense capital expense of modern semiconductor fabs requires near-total utilization to be profitable. This makes the integrated device manufacturing (IDM) model, where a company like Intel designs and builds its own chips, financially precarious if its own products cannot fill the fab's capacity.

Historically, general-purpose computing doubled transistor density at flat pricing every year, keeping baseline infrastructure spend relatively contained. Now that additional transistor capacity requires proportional capital outlays, growing digital consumption naturally expands the semiconductor sector into the largest industry in the world.

The manufacturing requirements for AI compute are staggering. Producing the advanced logic and memory wafers for just one gigawatt of data center capacity requires the output of approximately three and a half EUV lithography machines from ASML, representing over $1.2 billion in capital equipment.

While energy is a concern, the highly consolidated semiconductor supply chain, with TSMC controlling 90% of advanced nodes and relying on a single EUV machine supplier (ASML), creates a more immediate and inelastic bottleneck for AI hardware expansion than energy production.

Eighty percent fixed costs make fab profitability depend on utilization | RiffOn